What does sub pixel radius sharpening mean in this review of the Pentax K5IIs mean?How does it differ from regular sharpening?

Now this is the same raw file, with a 301% .3 sub pixel radius sharpening applied (sub radius is the best kind of sharpening for fine detail). The detail is simply ridiculous, and this is not the sharpest lens that can be used on the K5IIs, although...

  • Short story: it's the same thing. Slightly longer story: it's the same thing, just called something else by someone who doesn't understand what the term "radius" refers to. A more detailed explanation will have to wait for tonight.
    – JenSCDC
    Sep 19, 2014 at 17:12
  • Arghh!!! My explanation just isn't coming together. Before I take another shot, could someone confirm that the definition of "radius" that I should use is the the effective radius of the kernel? ISTR that different programs use different definitions.
    – JenSCDC
    Sep 19, 2014 at 23:23

2 Answers 2


Sharpening at that level is usually done with a deconvolution sharpener, such as the Smart Sharpen filter in Photoshop/Elements (or at higher radii, using a plugin like Topaz Labs InFocus). Unlike the usual sharpening tools like Unsharp Mask, which are just edge/contrast enhancers, deconvolution sharpeners attempt to compute what the data would have looked like with some amount of blur removed (that is, if the lens were perfect).

When you apply deconvolution sharpening at higher radii, you can extract detail from pictures that are badly out of focus or remove some amount of motion blur, usually at the expense of producing some ugly artifacts in parts of the picture that aren't the subject of interest (both bokeh regions and things that might have been in focus already will suffer).

When you use deconvolution sharpening at a low radius, especially under 1 pixel, it can largely overcome the effects of antialiasing (optical low pass) filters, Bayer-type colour filter arrays, and minor imperfections in the lens. It won't make things "pop" the way that unsharp masking will, since it's not so much about enhancing existing contrast as determining where the areas of contrast would have been in a more perfect optical and image recording system. You would still need to perform the same level of creative and output sharpening to get the "same" picture out of your workflow (for some value of "same"), but the final result can look like you've significantly upgraded your camera or lens.

(Higher radius deconvolution, at or just above one pixel — the sort of thing you might use a plugin for — works especially well for older cameras with less than 10MP resolution and the heavy AA filters of the period, as well as on moderately upscaled images, such as you get when you choose "the resolution higher than your camera actually provides" option in the Adobe Camera Raw or Lightroom output options.)


In USM (unsharp masking) radius is not the size of the kernel, but related to sigma. A good rule of thumb is that actually visible radius of the effect is close to radius * 3.3. So USM with the "radius" of 0.3 is pixel-level sharpening.

  • I think your first sentence could be reworked. Right now it's a bit misleading, not to mention that you should define what a kernel is.
    – JenSCDC
    Sep 20, 2014 at 3:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.